Commenters discussed the 2022 "Vectorized and performance-portable Quicksort" write-up and its claim of being the first portable vectorized Quicksort. Minitech explained the core idea: using a compress-store instruction (or emulating it with permutes) to partition, and porting that approach across multiple ISAs with Highway's portable SIMD. ZX41ZdbW and sciencesama noted newer algorithms (driftsort, ipnsort) and practical integrations like ClickHouse. Mixologic and atiedebee questioned whether one should rely on a 3,000-line portable SIMD library versus asking an LLM to reimplement code, with atiedebee preferring battle-tested libraries and mixologic raising the possibility of LLM assistance. Starcast2026 complained about a large 9 MB image in the post.
Opinion split on usefulness and aesthetics. Bee_rider and rodrigosetti made quips about Quicksort’s name and lack of elegance, while thesz and teiferer defended quicksort’s simplicity and in-place cache-friendly properties. Brrrrrm suggested radix sort often wins for numeric keys, and kg pointed to histogramming as a practical use for sorted samples. RVZ joked about unrealistic interview tasks, and shadowgovt lamented a software discoverability problem while praising LLMs’ ability to synthesize disparate solutions. Mcdonje and others implied adoption has been gradual.
Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.